CRYSTAL DEFECTS AND CONTAMINANTS gold density could have been adjusted to match the calculated and measured Hall coefficients, it was found that the gold density changes needed to accomplish this were somewhat larger than the estimated uncertainty of the gold density determinations. The fit of the calculated curve to the experimental data at higher and lower temperatures was examined for different values of the gold acceptor energy to determine which value gave the best fit. The data at the low temperature end, where the resistivity is very high, were not given much weight in selecting the best value for the gold acceptor energy. It is thought that surface conduction or other experimental problems caused the data points to deviate downward from the true bulk values. Data from measurements on different days on these and other high resistivity specimens often showed shifts where the magnitude of the change increased as the resistivity increased or as the temperature decreased. The points plotted for each specimen are the highest set of values obtained as they are considered to be most representative of the bulk values. In each case, it is evident that the experimental results fit the shape of the curve calculated with the acceptor level 0.535 eV below the conduction band edge better than they fit either of the other two curves. Although it is clear fixing the acceptor to the valence band e does not give a satisfactory fit, it is :: sible that allowing the acceptor level :: maintain its same relative position in the forbidden gap as the temperature changed might give as good or better fits. Numerous values of the activation energy the gold acceptor have been reported in th literature. Tasch and Sah [22] and Sah en al. [10] concluded from emission rate da as a function of temperature that the le is 0.59 eV above the valence band. Par and Johnson [23], also from emission rate measurements, concluded that the level is 0.72 ± 0.01 eV above the valence band. Collins et al. [13] concluded from Hall resistivity data that the level is 0.62: 0.02 eV above the valence band. Detaile: agreement with the present work depends the temperature variation of the acceptor level with respect to the band edges. : establish the temperature variation, an proved model for carrier mobility is required so that more accurate fits can be made to the data near the Hall inversic Further analysis of the Hall effect data being deferred pending investigation of cedures for calculating electron and hole mobilities. (W. R. Thu FILM CHARACTERIZATION 5. OXIDE FILM X-Ray Photoelectron Spectroscopy angular dependence of x-ray photoelecspectra (NBS Spec. Publ. 400-4, p. 42) investigated to study its applicability he determination of depth distributions urface impurities and thicknesses of surface films. A wide spectral survey of air stabilized silicon (which has a ective oxide coating resulting from room erature aging in air) revealed three ificant elements: silicon, oxygen and on. The spectra associated with these ents are shown in figure 10. relative area of each peak as a function lectron emission angle, e, is shown in re 11. The similar responses of the er kinetic energy silicon peak, labeled , and the oxygen peak supports the asso:ion of the Si02 peak with an oxide of con. The smaller carbon peak, labeled displays the same behavior and may be to carbon bonded to oxygen in the silioxide layer. The essentially flat angudependence of these peaks also suggests = the oxide layer is thicker than the tron escape depth. The higher energy pon peak, CÃ, has maximum intensity for tron emission parallel to the surface. s behavior may be expected of a thin erficial carbon layer. In contrast, the mer energy silicon peak, which is associI with elemental silicon is largest for tron emission perpendicular to the sure. This is typical of a substance Figure 10. X-ray photoelectron spectra of carbon, silicon, and oxygen from an airstabilized silicon specimen. Figure 11. 40 60 80 100 -20 O 20 ELECTRON EMISSION ANGLE (deg, Relative areas of x-ray photoelectron spectral peaks as a function of elect emission angle. (0: oxygen; : total silicon; : total carbon; □: Si; o: Si02; 0: The inset depicts the geometrical arrangement and the electron emission angle, OXIDE FILM CHARACTERIZATION lso observed to rise. A mass spectrometer onitored the vacuum chamber gases during eating. At low temperatures the primary as constituent was hydrogen, but as the emperature was raised the carbon monoxide ontent began to increase. At the highest emperatures, the mass spectrum became exremely complex. he drop in the total carbon photoelectron ignal is interpreted as a loss of this eleent from the specimen surface. The near onstancy of the elemental silicon to the 102 signal ratio indicates that practically o additional oxide was formed or removed rom the specimen surface. The oxygen inensity variation is more puzzling. The inetic energies and, correspondingly, the ean free paths for the oxygen electrons are dess than for the silicon electrons. Thereore, the removal of a surface contamination ayer should have a more pronounced effect pon the oxygen photoelectron emission. If 11 the oxygen were associated with the 102, then the oxygen peak should increase o a greater degree than the silicon peak. ince this was not the case, some of the xygen may have been removed with the caron, which suggests that not all the oxygen As associated with the silicon dioxide. and oxygen photoemission intensities. Whether all the excess (non-SiO2) oxygen was removed is not known, but it does seem clear that at least some fraction of the oxygencontaining adlayer has been removed. In view of the above off-gassing data, one might wonder whether the angular distribution of photoelectrons from non-Si02 oxygen and Si02 silicon should actually track one another as in figure 11. The expl lanation may lie with the orientation of the adsorbed molecules. The oxygen end of the molecule is certain to be more polar than the hydrogen fraction and probably rests against the SiO2 layer. These two oxygen signals would then contribute to a single oxygen signal, similar to that in figure 10b, which would track the SiO2 angular variation. (N. Erickson*, J. T. Yates, Jr.*, T. E. Madey, and A. G. Lieberman) 5.2. Comparative Study of Surface Analysis Techniques A series of measurements has been initiated to compare various electron, ion and photon beam technologies for the determination of the depth profiles of impurities in silicon and silicon dioxide. In these preliminary measurements, specimens of boron and zinc implanted silicon and of aluminum and sodium implanted silicon dioxide are being used to provide, in most cases, a reasonably well known density of impurity at a reasonably well defined location. The specimens were selected from materials contributed by various semiconductor processing houses. The properties of the specimens are summarized in table 2. For each specimen the implantation angle (measured between the incident beam and the normal to the specimen surface), Specimens for the Comparative Study of Surface Analysis Techniques OXIDE FILM CHARACTERIZATION the implantation energy, and the total dose Samples of these specimens are being subjected to secondary ion mass analysis (SIMS) [26], low and high energy ion scattering spectrometry (ISS [27] and RBS' * spectively), Auger electron spectroscopy [28], re(AES) [29], x-ray photoemission spectroscopy (XPS or ESCA) [30] and other techniques in laboratories associated with instrument manufacture, semiconductor analysis, or semiconductor device production. mens are expected to exercise the capabiliThe specities of the various measurement techniques and to provide illustrations of their different realms of applicability. Some characteristics of the results can be anticipated on the basis of the physical processes involved. For detecting boron in a silicon matrix, it is expected that SIMS will demonstrate certain advantages over other techniques. The sensitivity of SIMS to small quantities of material and the virtual impossibility that a molecular ion having almost the boron mass may form during sputtering together imply that the density profile of boron in the boron implanted silicon specimen should be easily measured, even with a low resolution SIMS instrument. On the other hand, the low atomic mass of boron relative to silicon makes boron a difficult element to detect by an ion scattering technique at any energy. In addition, due to the low atomic number of boron, the cruss-sections for x-ray photoemission and Auger electron emission are both very small, even though the Auger process is considerably more probable than the occurance of x-ray photoemission. Zinc The zinc implanted silicon specimen presents a reasonable cross section for XPS and A processes. However, while the sputter for SIMS may be high, the enormous peak sity of zinc atoms is nearly equal to the density of silicon atoms in the crystal severe matrix effects should be anticipata. in the SIMS measurement. Chemical shifts. the XPS and AES spectra may also arise te the depth of maximum implantation. Any measurement technique resulting in lized heating or charging of the sodium planted silicon dioxide specimen is capabl of displacing the sodium distribution [ This is true to varying degrees for all techniques being discussed. Furthermore, the sensitivities of XPS and AES to sodi are low, and the mass ratio of sodium to silicon or silicon dioxide is unfavorable for ion scattering to be useful. Channe cannot be used to enhance the Rutherfor: backscattering spectrum because the Si0: film is amorphous. From the measurements viewpoint, the al num implanted silicon dioxide specimen is interesting because aluminum neighbors * silicon in the periodic table and interfer ences may be expected in some cases. For example, even under optimum conditions the mass resolution M/AM is no better than for ISS measurements [27]. In addition, very shallow depth of the aluminum impla tion provides a test for the depth resol.tion of the various techniques. A trade off exists with respect to the intensity of the analyzing beam and the mer surement time for a given signal strength. By reducing the beam intensity and integr ing the signal over a longer period of tim the presence of elusive elements on a surface can often be detected by less sensiti techniques. The duration of a measurement is limited by the degree of sample destres tion that can be tolerated during an anal sis and by the patience of the investigat Thus while the sensitivity of photoemissi Auger emission and ion scattering techniq may not be as great as that of SIMS, these techniques are far less destructive to the specimen surface and the detection limit ca be enhanced by prolonging the measurement interval. (A. G. Lieberma |